As you will remember AMD introduced its Ryzen AI 300 chips at the beginning of Junea few weeks before the arrival of summer. And likewise, weeks before the end of the mentioned season, it confirmed its arrival on the market in more portable devices that are added to the Copilot+ PC platform and that we could see for the most part at IFA 2024. , the Berlin technology meeting that marks the return to activity after the inevitable summer a break which, however, is a little less each year than the previous one.

As is inevitable these days, the focus of this new generation of AMD APUs put into his abilities for AIwhich in this case are very directly related to the platform defined by Microsoft and which emphasizes the need for computing power that it considers necessary to easily use the features of Windows 11, depends exclusively on the NPU of integrated statements. Now, not everything is Copilot, nor should AI-based features depend solely on the NPU. What’s more, AI isn’t everything, even if it’s starting to look different.

The most recent example can be found with the recent announcement from AMD just confirmed the arrival of Fluid Motion Frames 2 (AFMF 2) and Variable Graphics Memory (VGM) to new notebooks equipped with APU Ryzen AI 300, two technologies that promise a substantial improvement in the user experience of these devices (even those based on other integrated devices compatible with them), especially with regard to the intensive use of their graphics part. , i.e. mainly to play.

AMD Fluid Motion Frames 2

AMD Fluid Motion Frames 2 is, as you already know, the second generation of AMD technology, which marks the reaction of this technology to NVIDIA DLSS 3 Frame Generation, that is It is able to generate “medium” images, exclusively with the GPU and the use of artificial intelligence, which are interspersed between those generated in a traditional way, so they also count on the CPU. Capable of utilizing the specialized functions of graphics adapters, these image generation technologies help significantly reduce the known CPU bottleneck.

At the end of July, AMD released a technical preview of AFMF 2, and from what we could see at the time, The Santa Clara guys seem to be hitting the gas after a somewhat improved debut of this technology in its first generation. As we told you back then, the announcement of DLSS 3 caught AMD on the wrong foot, forcing the tech company to make the announcement more than a year after AFMF came to users. And yet it was evident with his debut that he still has quite a bit of work to do.

Now, with AFMF 2 now available in Preview for users of APU-based systems, the Ryzen AI 300, all indications are that AMD has caught onand is therefore now in a position to offer an alternative to NVIDIA’s proposal. Adding that, like its first generation, Fluid Motion Frames 2 has a much wider scope and, according to AMD, is much easier to implement.

As a reminder, these were the main new features that AMD says will arrive with AFMF 2:

• Improvements optimized for artificial intelligence – AFMF 2 provides significant improvements in image generation with AI optimization by developing an updated algorithm, resulting in smoother image generation. Additionally, to allow more control over how AFMF works with your games, AFMF 2 introduces two new modes where the optimal settings are activated automatically, but advanced users can adjust them to their preferences.
• Performance improvement – AFMF 2 features a new “Performance” mode that reduces the overhead of AFMF 2, helping to make high frame rate gaming experiences achievable on a wider range of devices. This is especially beneficial when using AFMF 2 with integrated graphics cards, and is now the default “Auto” setting on AMD Ryzen processors that support Radeon graphics.
• Lower latency frame generation – experience significant improvements in latency reduction added by image generation with the latest version of AFMF. These improvements apply across the board when using AFMF 2, regardless of configuration mode, resolution or hardware used.
• More updates and improvements – AFMF 2 now supports borderless full screen mode when using AMD Radeon™ RX 7000 and Radeon 700M graphics cards. Additionally, AFMF 2 now supports games that use Vulkan and OpenGL, further expanding the already huge number of games that can be used with AFMF. Additionally, interoperability with AMD Radeon Chill has been enabled.

Variable graphics memory

Less well known, but no less important, Variable graphics memory represents a significant evolution of the shared memory technology previously present in AMD integrated devices.. As you can imagine, both have a common root, which is to “steal” a part of the system RAM to add it to the system VRAM, thereby increasing the system’s capabilities when performing tasks (and by tasks, of course, I include gaming) especially graphics-intensive and/or with system VRAM.

So in this case we find a proposition which allows you to “convert” up to 75% of RAM to graphics memoryand as its name suggests, it allows us to make this type of adjustment dynamically depending on the needs of the game we’re going to use (or an application, in some cases I can think of). And this is especially useful if we are talking about laptops with a graphics adapter integrated in the APU, such as systems with Ryzen AI 300, which tend to have low VRAM and therefore do not cover the minimum of many games.

Another use in which VGM can be particularly practical is indeed the above to applications that use local AI modelsand as you already know, these types of applications (or rather the models they use) tend to be very demanding on the amount of graphics memory in the system. Thanks to this technology, which is already available in the Ryzen AI 300, users will be able to bypass this limitation.

These two functions, as I indicated at the beginning, are available in va a preliminary version included in the Adrenalin special editionwhich you can download here. However, keep in mind that while we can expect them to be quite polished, we’re not yet on the final versions of these technologies, so it’s possible that they won’t work 100% as expected.